JP2009093383A - Method and device for controlling multicore processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a multi-core processor control method and device that select and operate an appropriate number of cores matching an overall operation state of a processor. <P>SOLUTION: In a multi-core processor having a plurality of cores each independently performing operations on one processor, software in a specific core calculates the operating rate of threads or tasks of each core within a predetermined time from a total operating time or operating frequency within the predetermined time, and sums up such operating rates to calculate the operating rate of all the cores. The number of cores that should be operated is determined on a preset table according to the overall operating rate. The number of cores has a hysteresis characteristic as the overall operating rate increases and decreases. The cores that should be operated are selected on the preset table according to the determined number of cores. As for exception handling, all the cores are operated for a predetermined time, after which the original process is resumed if a check finds the exception handling resolved. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a control method for reducing power consumption in a unit equipped with a processor having a multi-core, and an apparatus for implementing the method.

  In a general microprocessor that has been used conventionally, a set of processor cores that operate as a single component, in which a command issuing unit, a calculation unit, and the like are combined, is incorporated in a package. In order to improve the performance using such a single processor, a clock-up method for increasing the clock frequency has been used conventionally. However, the increase in power consumption and heat generation associated with clock-up are major bottlenecks, and the development of other methods has been demanded.

  As a countermeasure, a multiprocessor technology has been developed and widely used in which a plurality of processors are connected so that they can be operated in parallel so that processing can be performed efficiently. However, if a plurality of processors are simply connected so as to be able to operate in parallel, there is a problem that if the number of processors increases, the size of the processors must be increased, and the use range of the processors is limited. As a countermeasure, a multi-core processor in which a plurality of processor cores capable of operating independently from each other has been developed and widely used due to recent progress in high-density LSI manufacturing technology.

  In a multi-core processor, each processor core is basically independent, so that each processor core can operate without being influenced by other processor cores. The principle of a multi-core processor is almost the same as that of a multi-processor. Processing can be shared by the processor core, and the performance can be improved accordingly. Therefore, if the number of cores mounted on the multi-core processor is increased, the performance will be increased accordingly.

  For this reason, the number of multi-core processors, such as a dual-core processor having two cores and a quad-core processor having four cores, is gradually increasing. Such a multi-core processor is treated as a plurality of microprocessors from the OS, and the operation feeling is almost the same as that of the multi-processor configuration. Therefore, users and programmers should work without being particularly aware of the operation on the multi-core processor. Therefore, it is attracting attention as an easy-to-use processor.

In a multiprocessor system composed of a plurality of processors, when a large amount of processing capacity is not required, a technology for monitoring the usage amount of each processor, identifying a processor with a low usage amount, and stopping or suspending it is disclosed in JP-A-11. -202988.
JP 11-202988 A

  As described above, in a multi-core processor, a plurality of processors are mounted on one processor, which is small as a whole and is easy to handle because the processors are integrated into one, and is gradually becoming cheaper. Is also about to be widely used. However, when using this high-performance multi-core processor, for example, when the multi-core processor is a quad-core processor having four cores in one processor, it is not always necessary to use four cores.

  In other words, when trying to improve the performance of a specific electrical device, when a specific function is performed, a large amount of processing is required in a short time, so that a quad core processor having four cores is mounted. Especially, when operating a function that performs large-capacity processing, the four cores are fully operated to cope with a desired function, but when performing other functions, this is the case. Often does not require high-speed processing.

  However, the conventional multi-core processor is always in a standby state to perform some processing even when there is no task to be processed at present or when thread processing in one task is not performed. I need power. Such unnecessary power consumption is a problem even in general household devices, but power is supplied from batteries that supply power to many devices, especially portable devices that operate on batteries and lighting devices. For example, it is a heavy burden for a vehicle-mounted device, and countermeasures are required. Furthermore, since the amount of heat generated is large when the power consumption is high, measures for suppressing various types of heat generation and measures for the generated heat are required, which causes an increase in cost as a whole.

  In the technology described in Patent Document 1 cited as the prior art, when a large amount of processing capacity is not required in a multiprocessor system, the usage amount of each processor is monitored, and a processor with a low usage amount is identified and stopped. Although a technique for suspending is disclosed, it merely restricts the operation of a specific processor, and does not necessarily control the operation of the processor efficiently from the viewpoint of the entire processor.

  Therefore, according to the present invention, in a multi-core processor having a plurality of independently operable processor cores, when the total amount of processing performed by the processor is small, an appropriate number of cores are operated as a whole, and the entire operation is performed. A multi-core processor control method that reduces power consumption, suppresses heat generation, prevents heat generation, and reduces costs required for heat countermeasures generated by preventing the operation of cores that do not need to be operated, and its The main object is to provide an apparatus for carrying out the method.

  In order to solve the above problems, a multi-core processor control method according to the present invention is a multi-core processor control method including a plurality of cores that perform arithmetic processing independently on a single processor. The rate is calculated, the operating rates of all the cores are totaled, the number of cores operating according to the total operating rate is determined, and the cores operating according to the determined number of cores are selected.

  In another multi-core processor control method according to the present invention, in the multi-core processor control method, the operating rate of each core is an operating time within a predetermined time, and the total operating time of each core It is characterized by calculating the operating rate of

  Another multi-core processor control method according to the present invention is the multi-core processor control method, wherein the operating rate of each core is the number of operations within a predetermined time, and the total number of operations of each core It is characterized by calculating the operating rate of

  Another multi-core processor control method according to the present invention is characterized in that, in the multi-core processor control method, the operation rate of each core is calculated for a thread or task performed in each core.

  Further, another multi-core processor control method according to the present invention, in the multi-core processor control method, each processing of calculating the operating rate of each core and all cores, determining the number of operating cores, determining the operating cores, It is performed by a specific core among the plurality of cores.

  According to another multi-core processor control method of the present invention, in the multi-core processor control method, when the operating core is selected, the hysteresis characteristic is obtained by changing the operating number between when the total operating rate increases and when the operating rate decreases. It is characterized by having.

  Further, another multi-core processor control method according to the present invention is characterized in that, in the multi-core processor control method, when selecting the plurality of cores, a specific core predetermined according to the total operation rate is selected.

  In addition, another multi-core processor control method according to the present invention is characterized in that, in the multi-core processor control method, when selecting the plurality of cores, the multi-core processor control method is changed according to a predetermined method according to the total operating rate.

  Another multi-core processor control method according to the present invention is characterized in that, in the multi-core processor control method, the operating core is selected at random.

  Another multi-core processor control method according to the present invention is characterized in that, in the multi-core processor control method, when selecting the plurality of cores, all cores are selected when occurrence of exception processing is detected.

  In another multi-core processor control method according to the present invention, in the multi-core processor control method, when it is detected that an exception processing state has been resolved after a predetermined time has elapsed after selecting all the cores, The selection process is restarted.

  According to another multicore processor control method of the present invention, in the multicore processor control method, the exception processing is performed when it is detected that the operation of the multicore processor control method is not properly performed. Features.

  In addition, in order to solve the above problems, a multi-core processor control device according to the present invention is a multi-core processor control device including a plurality of cores that perform arithmetic processing independently on a single processor. A core operating rate calculating unit for calculating the operating rate of the processor, a total operating rate calculating unit for calculating the operating rate of the entire processor by summing up the operating rates of the respective cores calculated by the core operating rate calculating unit, and an overall operating rate of the processor Operating core number storage means for correspondingly storing the relationship between the rate and the number of operating cores, operating core number determination means for determining the operating number from the movable core number storage means based on the overall operating rate of the processor, and the movable core number It is characterized by comprising operating core selecting means for selecting a core that operates in accordance with the number of cores determined by the determining means.

  In the multi-core processor control device according to the present invention, in the multi-core processor control device, the core operation rate calculation means calculates a core operation rate for each core based on an operation time within a predetermined time, and the overall operation The rate calculating means is characterized in that the total operating rate is calculated by summing up the operating time of each core.

  Further, in another multi-core processor control device according to the present invention, in the multi-core processor control device, the core operation rate calculation means calculates the total operation rate by totaling the number of operations within a predetermined time, and the total operation rate The rate calculating means is characterized in that the total operating rate is calculated by summing up the number of operating times of each core.

  Further, another multi-core processor control device according to the present invention is characterized in that, in the multi-core processor control device, the core operation rate calculation means performs a thread or a task performed in each core.

  Another multi-core processor control device according to the present invention is the multi-core processor control device, wherein the calculation means, storage means, determination means, and selection means are provided in a specific core among the plurality of cores. Features.

  Further, another multi-core processor control device according to the present invention comprises a hysteresis characteristic in the multi-core processor control device, wherein the movable core selecting means has a hysteresis characteristic by making the number of operations different between when the overall operation rate is increased and when it is decreased. It is characterized by that.

  Another multi-core processor control device according to the present invention is characterized in that, in the multi-core processor control device, the movable core selection means selects a specific core predetermined according to the total operation rate.

  Further, another multi-core processor control device according to the present invention comprises exception processing occurrence detection means for detecting an exception processing occurrence state of the multi-core processor in the multi-core processor control device, wherein the exception processing occurrence detection means When it is detected that the number of movable cores is determined, the movable core number determining means determines the movable of all the cores.

  Since the present invention is configured as described above, in a multi-core processor having a plurality of processor cores that can operate independently, when the total processing amount processed by this processor is small, an appropriate number of cores as a whole is selected. Necessary cores that do not need to be operated can be reduced by reducing power consumption and reducing power consumption by reducing overall power consumption. A multi-core processor control method capable of reducing cost

  The present invention aims at operating only an appropriate number of cores in accordance with the operating state of a multi-core processor, in a control method for a multi-core processor having a plurality of cores that perform arithmetic processing independently on a single processor. The operating rate of each core is calculated and the operating rates of all cores are totaled. The number of operating cores is determined according to the total operating rate, and the operating core is selected according to the determined number of cores. It was realized by doing.

  Embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a functional block diagram of a multi-core processor according to the present invention, and shows an example in which one processor includes four cores (Core 0 to 3). In the example shown in the figure, one task is divided into a number of threads, and each thread is assigned to each core by the thread scheduler 1 to perform processing in order to increase processing speed.

  Accordingly, in the example of FIG. 1, there are a number of tasks such as the first task 3, the second task 4... In the task list 2, for example, a number of threads exist for the first task 1, and these threads are thread schedulers. 1 is processed by cores 0-3. At that time, the thread scheduler 1 performs thread allocation processing based on the thread information in the task list.

  Therefore, in the illustrated example, the cores 0 to 3 are the core 0 thread 5 that is the thread that is processed by the core 0 for the first task 3, the core 1 thread 6 that is the thread that is processed by the core 1, and the core 2 In this case, there are a core 2 thread 7 that is a thread that is to be processed in the above and a core 3 thread 8 that is a thread that is to be processed in the core 3. When performing processing of task 1 in this way, core 0 processes core 0 thread 5 for first task 3 shown in task list 2, and similarly core 1 processes core 1 thread 6 and core 2 Then, the core 2 thread 7 is processed, and the core 3 processes the core 3 thread 8. As described above, one task is divided into a large number of threads, and parallel processing is performed by four cores in the figure. As a result, desired processing can be performed as a multi-core processor.

  In such a multi-core processor, one of the four cores 0 to 3 is set as a basic core that is a movable control core. The illustrated example shows an example in which the core 0 is the basic core, and the core 0 includes an operating core selection processing unit 9 that performs processing to select an operating core among the cores 0 to 3. The processing performed by the operating core selection processing unit 9 will be described in detail later with reference to FIGS. 2 to 5, but the operating core selection processing unit 9 shown in FIG. Based on the calculation result of the calculation unit 10, the total operation rate based on the calculation result of the per-core operation rate calculation unit, and the operation core storage unit 15 based on the calculation result of the total operation rate calculation unit Based on the number of operating cores in the immediately preceding process stored in step 1, the data is read from the operating rate / active core number correspondence table 13, and the operating core number determining unit 12 that determines the number of operating cores is determined by the operating core number determining unit. The example provided with the active core selection part 14 which selects an active core based on the number of active cores and the data of the last active core as needed is shown.

  In the multi-core processor composed of the functional blocks as described above, the present invention can be operated in sequence according to the operation flow as shown in FIGS. FIG. 2 shows an operation flow of the operation rate calculation process for each core, which is basic data for the control of the present invention. In the illustrated example, the operation rate measurement process is first performed as the process for each core. It is determined whether or not to start (step S11).

  That is, when measuring the operating rate for each core, for example, as shown in FIG. 4, during the predetermined time T, the time during which each core is operating, that is, the predetermined time indicated as ts-tf in the figure. In order to measure the operating rate of each core at time T, ts is set as the measurement process start time and tf is set as the end time. Therefore, in step S11 of the operation flow of FIG. 2, it is determined whether or not this time ts has been reached. Note that the actual processing can be executed without detecting the time by issuing a processing start instruction and an end instruction with software provided in advance in the core 0, for example.

  In the example of FIG. 2, when there is a process start instruction as described above in step S11 and it is determined that the operation rate measurement process is started, the operation rate measurement unit time has ended before the next process is performed. In addition, it is determined whether or not the operation rate measurement unit time has ended (step S12). When the operation rate measurement unit time has ended, the process proceeds to step S21 to be described later, and when it has not ended, the process proceeds to step S13. Note that when there is no instruction to start the operation rate measurement process yet in step S11, the apparatus stands by in this state. In the example of FIG. 4, it is determined that the core 0 has started the processing of the thread A after the time ts when the operation rate measurement processing is started. If the processing of the thread A has not yet been started in step S13, the process returns to step S12, and if the operation rate measurement unit time has not ended, this operation is repeated and waits.

  At this time, it is determined whether or not the operation rate measurement process is started in step S11, it is detected that the start time has been reached, and immediately after it is determined in step S12 that the operation rate measurement unit time has not ended. In step S13, it is determined whether or not thread processing has started, and when it is determined that the thread processing has started at this time, the thread has started thread processing before the operation rate measurement start time ts, In such a case, the measurement of the operation time starts from the time ts during the thread processing.

  When it is determined in step S13 that the thread process has started, the start time of the process is read from the timer. In the example of FIG. 4, when the processing of the thread A is started in the core 0, the time ta1 at that time is read from the timer. Next, in the illustrated example, it is determined whether or not the operation rate measurement unit time has ended (step S15). That is, in this mobility measurement process, the time T from the time ts to tf is set as the operation rate measurement processing unit time, and in the step S15, it is determined whether or not the operation rate measurement unit time T has ended. ing. The end of the operation rate measurement unit time can be determined by whether or not a measurement end instruction is issued by the control device.

  If it is determined in step S15 that the operation rate measurement processing unit time has not yet ended, the process proceeds to step S16, and it is determined whether the thread process has ended (step S16). In the example of FIG. 4, it is determined whether or not the thread A started at time ta1 has ended. Here, when it is determined that the processing of the thread A has not yet ended, the process returns to step S15 again to determine whether or not the operation rate measurement unit time T has ended, and these operations are repeated.

  As described above, when the thread processing is completed while the operation rate measurement unit time has not yet ended, the thread processing end time is read from the timer (step S17). In the example of FIG. 4, the processing of the thread A of the core 0 ends at time ta2, and this time ta2 is read.

  Next, the operating time is calculated from the thread processing start time ta1 read from the timer in step S14 and the thread end time ta2 read from the timer in step S17, and the value is stored (step S18). In this process, the processing time of the thread A, that is, the operating time T1 of the thread A in the core 0 can be calculated by (ta2-ta1). Thereafter, it is determined again whether or not the operation rate measurement unit time has ended (step S19), and if not ended, the process returns to step S12 again and goes through a similar process that mainly performs a determination process for the other processes. Then, after waiting for the start of the next thread processing in this core, the same operation is repeated.

  If it is determined in step S15 that the operation rate measurement unit time has ended, the process proceeds to step S22 in the illustrated example, and the time at which the operation rate measurement unit time ends is read from the timer.

  In the example of FIG. 4, the processing of the thread B after the processing of the thread A is performed, and the processing of the thread B starts at time tb1 and ends at time tb2. In this process, T2 is obtained from the operating time (tb2-tb1) as described above. In the core 0, the processing of the thread C is further performed during the operation rate measurement unit time T. In this thread C, the processing starts at the time tc1, and ends at the time tc2, and the result (tc2-tc1) To obtain the operating time T3 of the thread C. These processes can also be performed by repeating the processes from step S13 to step S19 in FIG.

  After determining that a new thread process is started in step S13 in this process, when determining in step S15 whether or not the operation rate measurement unit time has ended, In such a case, the process proceeds to step S22. In the illustrated example, the end time of the operation rate measurement unit time is read from the timer. However, since the end time tf of the operation rate measurement unit time is set in advance, the operation rate measurement unit end time may be acquired by directly reading this set value.

  Next, in this thread process, the operation time is calculated and stored in the same manner as described above from the thread process start time and the end time of the operation rate measurement unit time. In step S18 and step S19, when the thread operation is not performed, when the operation rate measurement unit time is completed, the processing of step S22 and step S23 is performed, and each thread within the operation rate measurement unit time is processed. The operating hours are totaled (step S20). When it is determined in step S19 that the operation rate measurement unit time has ended, the total calculation is performed in step S20 as described above, and at the same time, the process returns to step S11 to determine whether the next operation rate measurement unit time has started. Is determined. In many cases, since the operation rate measurement unit time is continuous, in this case, immediately after step S19, the process proceeds from step S11 to step S13, and the above operation is repeated.

  In the process of totaling the operating time for each thread within the operating rate measurement unit time, in the example of FIG. 4, in core 0, T1 of thread A, T2 of thread B, and T3 of thread C are summed (T1 + T2 + T3). ). In core 1, only T4 of thread D is provided, and in core 2, T5 of thread E and T6 of thread F are summed to calculate (T5 + T6). Similarly, in core 3, T7 of thread G and T8 of thread H are summed to calculate (T7 + T8).

  Next, in step S21, the operation rate is calculated within the operation rate measurement unit time, and the calculated value is output (step S24). In this process, as shown in FIG. 4, in the core 0, the operating rate η0 is calculated as [η0 = (T1 + T2 + T3) / T × 100], and in the core 1, the operating rate η1 is [η1 = T4 / T × 100]. In the core 2, the operating rate η2 is calculated as [η2 = (T5 + T6) / T × 100], and in the core 3, the operating rate η3 is calculated as [η3 = (T7 + T8) / T × 100]. Ask for.

  As a result of obtaining the operating rate for each core as basic data for microprocessor control according to the present invention by the above-described method, the operating core selection control processing by the control of the multi-core processor according to the present invention is as shown in FIG. Can be done. In the example of the operating core selection control process shown in FIG. 3, after the operating rate calculation process for each core as shown in FIG. 2 is first performed (step S31), that is, after the operating rate measurement unit time ends. After performing the operation rate calculation process for each core, the operation rate calculation values of the respective cores are aggregated to calculate the total operation rate (step S32).

  In calculating the total operating rate, the operating rate η0 of the core, the operating rate η1 of the core 1, the operating rate η2 of the core 2, and the operating rate η3 of the core 3 are totaled, as shown in FIG. The overall operating rate η is obtained by (η0 + η1 + η2 + η3). Therefore, for example, when all the cores are in operation during the operation rate measurement unit time T, η = 400, and when it is assumed that all the cores are not in operation, η = 0. However, in the present invention, the control processing according to the present invention is performed at least on a specific core, but the calculation processing amount is extremely small.

  Thereafter, in this processor, the number of operating cores is read from the table based on the number of operating cores immediately before the start of the operating rate calculation unit time, that is, the number of operating cores at the time of the previous operation and the overall operating rate calculated as described above (step S33). Various tables can be used as the table at this time. For example, an operating core number / core start / stop timing table as shown in FIG. 5 can be used.

  In the example of the operating core number / core start / stop timing table shown in FIG. 5, when the total operating rate η is 0 to 40%, the operating core number is 1, and as is clear from FIG. Only will be selected. If the overall operating rate gradually increases monotonically from this state, the number of cores is set to 2 when the overall operating rate η exceeds 80%, and the core 1 is activated in the illustrated example. Next, when the overall operating rate η exceeds 160%, the number of operating cores is set to 3, and the core 2 is also activated in the illustrated example. Further, when the total operating rate η exceeds 240%, the number of operating cores is set to 4, and the core 3 is activated in the illustrated example. That is, when the overall operating rate exceeds 240%, all the four cores are all activated for the first time.

  In this way, by setting the number of operating cores in each total operating rate state, it is possible to operate the number of operating cores according to the total processing amount, so the power consumption of this multi-core prosensor can be reduced. Can be reduced. In addition, since the core having a low operating rate is not simply stopped as in the prior art, the stopped core can be adjusted according to the operating rate of the entire processor, so that the optimal operating core can be controlled for the entire processor.

  When the overall operating rate gradually increases as described above, the operating core is controlled as described above, but conversely the number of operating cores is monotonically increased from 4 with the overall operating rate being 400%. When the operating rate decreases to 160 or less, the number of operating cores is set to 3, and the operation of the core 3 is stopped. Thus, in the example of the table of FIG. 5, for example, the state of operation with 4 operating cores is 240% when activated, but 160% after being activated once. Up to four are operating. The operation having such a hysteresis characteristic is similarly performed in the control of the number of cores at other overall operating rates as will be described later.

  As described above, when the overall operating rate gradually decreases and drops from 160% to 80%, the number of operating cores is set to 2, and in the illustrated example, the core 2 is stopped. Further, when the overall operating rate is reduced to 40%, the number of operating cores is set to 1, the core 1 is stopped, and the processing is performed only with the core 0 thereafter. Since this is set with a hysteresis characteristic that makes the number of core operations different between when the overall availability rises and when it falls, when the overall availability changes frequently, the core It is possible to prevent frequent start and stop operations of starting and stopping.

  Therefore, for example, the overall occupancy rate is (1) 10%-(2) 70%-(3) 90%-(4) 45%-(5) 200%-(6) 150%-(7) 300%-( 8) When changing in order of 70%, (1) 10% operates only core 0 with 1 active core, and (2) 70% thereafter operates only core 0 with 1 operating core, and (3) 90 % With 2 active cores, operating core 0 and core 1, and (4) 45% with 2 operating cores operating with core 0 and core 1, and then (5) 200 with operating cores Core 0, Core 1 and Core 2 are operated at 3, and then (6) 150%, the number of core operations remains unchanged, so Core 0, Core 1 and Core 2 are operated, and then (7) 300% Then, the number of operating cores is 4 and core 3 is also operating, and all cores are in operation, and then (8) 70% are operating In A few 2, core 4, so that the running of the core 0 and core 1 and further stop the core 3.

  Thus, for example, even when the same overall operation rate is 70%, when (1) changes from 10% to (2) 70%, only core 0 operates, and (7) changes from 300% to (8) 70%. Sometimes the core 0 and the core 1 are operated, and even when the overall operating rate is 45% lower than 70%, (3) when the operating rate is 90%, the core 0 and the core 1 are continuously operated. It is a table that determines the operating core in relation to the previous total operating rate, not just the data of the total operating rate.

  In step S33 of FIG. 3, after the number of operating cores is read based on the number of operating cores immediately before and the total operating rate calculated this time, the operating core is selected and stored in the example of FIG. 3 (step S34). ). That is, in the example of the table of FIG. 5, when the core 0 is the basic core for the cores 0 to 3 and the total operating rate requires two cores, the core 1 is always added, and further three operating cores are required. The example shows that core 2 is always activated at the time of the overall operation rate, and the core 3 is operated only at the overall operation rate where all the cores need to be operated. Even if it sets in this way, about the selection of the core to operate | move, it is not necessarily based on the said example, It can also drive by another method.

  In that case, for example, the order of selecting the cores to be operated according to the overall operating rate is the order of cores 0, 1, 2, and 3, in addition to the order of cores 0, 2, 3, and 1; The operation of the specific core corresponding to the preset overall operating rate, such as in the order of 3, 1, 2, etc., can be performed in various modes. In the above-described embodiment, as shown in FIG. 1, since the software for performing the processing of the present invention is incorporated in the core 0 and this operation processing is performed, this core is always operated as a basic core. Therefore, when this software is incorporated in the core 1 and the operation process is performed, the core 1 is always operated as a basic core.

  In the above example as well, even if the cores 0, 1, 2, and 3 are initially operated in accordance with the overall operating rate, only the basic core 0 is operating in the subsequent thread processing. At the right timing, for example, operate in the order of cores 0, 2, 3, 1 and then in the order of cores 0, 2, 1, 3 and in the order of cores 0, 3, 1, 2 and It may be preset to change the operation in various manners, such as in the order of 0, 3, 2, 1 and in the order of the cores 0, 1, 3, 2. At that time, when the basic core is set to the core 0, the processing load can be made uniform by controlling the operation of other cores as uniform as possible. In that case, it is also possible to change other than the basic core at random.

  In the example of the operation in FIG. 3, the operating core is selected as described above in step S34, and the selected core is stored. At this time, only the number of cores may be stored. Next, in the illustrated example, it is determined whether or not exception processing has occurred. That is, in the course of a series of thread processing of this processor, when a large amount of processing is urgently required, such as when a large amount of data processing is interrupted, exception processing occurs in step S35. In this case, the operation process in all the cores can be immediately performed (step S36).

  Various conditions can be set in advance as exception processing at this time. For example, when it is detected that the multi-core processor control is not operating properly, the above core selection processing is canceled. That is, it may be used as a safety measure for processing according to the present invention. In step S36, for example, when a large amount of data processing work is performed and all cores are operated, it is determined whether or not a predetermined time has passed (step S37). The operation is periodically repeated and the operation is periodically repeated. When a predetermined time has elapsed, the operation returns to step S31 and the operation is repeated. When it is determined in step S35 that exception processing has not yet occurred, a thread obtained by dividing one task into a plurality of cores is allocated to the core selected in step S34 by the thread scheduler 1 in FIG. 1 (step S38), and again. Returning to step S31, the same operation is repeated.

  In the above example, an example in which processing of a thread obtained by subdividing one task is processed by the thread scheduler for each core of the multi-core processor is shown. When the task is divided into tasks and each task is allocated to an arbitrary core by the task scheduler, it can be similarly operated by the above-described method. In this case, the thread scheduler of FIG. 1 becomes a task scheduler, and, for example, the core 0 task, the core 1 task, the core 2 task, and the core 3 task that are subdivided for the first task are allocated to each core. It will be.

  In the above example, when calculating the operating rate of each core, the operating time of each thread processed within a predetermined time is obtained, and the total is calculated to obtain the operating time of each core. Although an example is shown, the number of threads or tasks processed in a predetermined time may be obtained, and this processor may be configured as the operating rate of each core to control the processor by the above method. Thereby, it is possible to simplify the operation rate calculation process and reduce the processing load of the basic core performing this calculation.

It is a functional block diagram of the Example of this invention. It is an operation | movement flowchart of the operation rate calculation process for every core in the Example. It is an operation | movement flowchart of the active core selection control process in the Example. It is a figure which shows the operation rate calculation example for every core. It is a figure which shows the example of a table which selects the number of active cores at the time of operating core number determination.

Explanation of symbols

1 Thread scheduler
2 Task List 3 1st Task 4 2nd Task 5 Core 0 Thread 6 Core 1 Thread 7 Core 2 Thread 8 Core 3 Thread 9 Operating Core Selection Processing Unit 10 Core Operating Rate Calculation Unit 11 Total Operating Rate Calculation Unit 12 Number of Operating Cores Determining unit 13 Operating rate / number of operating cores correspondence table 14 Operating core selection unit 15 Operating core storage unit

Claims (20)

In a control method of a multi-core processor having a plurality of cores that perform arithmetic processing independently in one processor,
Calculate the utilization rate of each core within a given time and add up the utilization rate of all cores.
Determine the number of operating cores according to the total overall operating rate,
A multi-core processor control method, wherein a core to be operated is selected according to the determined number of cores. The operating rate of each core is an operating time within a predetermined time,
2. The multi-core processor control method according to claim 1, wherein the operating rate of all cores is calculated by adding up the operating time of each core. The operating rate of each core is the number of operations within a predetermined time,
The multicore processor control method according to claim 1, wherein the operation rate of all cores is calculated by summing up the number of operations of each core.   The multi-core processor control method according to claim 1, wherein the calculation of the operating rate of each core is performed for a thread or task performed in each core.   The calculation of the operating rate of each core and all cores, the determination of the number of operating cores, and the determination of the operating cores are performed by a specific core among the plurality of cores. Multi-core processor control method.   2. The multi-core processor control method according to claim 1, wherein when the operating core is selected, a hysteresis characteristic is provided by differentiating the operating number between when the total operating rate increases and when the operating rate decreases.   2. The multi-core processor control method according to claim 1, wherein, when selecting the plurality of cores, a specific core predetermined in accordance with the total operating rate is selected.   2. The multi-core processor control method according to claim 1, wherein the selection of the plurality of cores is changed by a predetermined method according to the total operating rate.   2. The method of controlling a multi-core processor according to claim 1, wherein the operating core is selected at random.   2. The multi-core processor control method according to claim 1, wherein in selecting the plurality of cores, all cores are selected when occurrence of exception processing is detected.   11. The multi-core processor control method according to claim 10, wherein, after selecting all the cores, the core selection process is resumed when it is detected that an exception processing state has been resolved after a predetermined time has elapsed.   11. The multi-core processor control method according to claim 10, wherein the exception processing is performed when it is detected that the operation of the multi-core processor control method according to claim 1 is not properly performed. In a control device for a multi-core processor having a plurality of cores that perform arithmetic processing independently on one processor,
Core utilization rate calculation means for calculating the utilization rate of each core within a predetermined time;
Total operating rate calculating means for calculating the operating rate of the entire processor by summing up the operating rates of each core calculated by the core operating rate calculating unit,
An operating core number storage means for storing the relationship between the overall operating rate of the processor and the number of operating cores in correspondence;
An operating core number determining means for determining the operating number from the movable core number storage means according to the overall operating rate of the processor;
A multi-core processor control device, comprising: an operating core selecting unit that selects a core that operates in accordance with the number of cores determined by the movable core number determining unit. The core utilization rate calculating means calculates the core utilization rate for each core based on the operation time within a predetermined time,
14. The multi-core processor control device according to claim 13, wherein the total operating rate calculating means calculates the total operating rate by adding up the operating times of the respective cores.   The core operating rate calculating means calculates the total operating rate by totaling the number of operating times within a predetermined time, and the overall operating rate calculating unit calculates the total operating rate by adding the operating times of each core. The multi-core processor control device according to claim 13.   The multi-core processor control device according to claim 13, wherein the core operation rate calculation unit performs the thread or task performed in each core.   14. The multi-core processor control device according to claim 13, wherein each of the calculation means, storage means, determination means, and selection means is provided in a specific core among the plurality of cores.   14. The multi-core processor control device according to claim 13, wherein the movable core selection means has a hysteresis characteristic by varying the number of operations depending on whether the total operation rate is increased or decreased.   The multi-core processor control device according to claim 13, wherein the movable core selection unit selects a specific core determined in advance according to the total operating rate. Exception processing occurrence detection means for detecting the exception processing occurrence state of the multi-core processor is provided,
14. The multi-core processor control device according to claim 13, wherein when the exception process occurrence detecting unit detects that an exception process occurs, the movable core number determining unit determines that all cores are movable.

Images